Low-rank adaptation with gating mechanisms in large language models, an improved method for fine-tuning: G-LoRA